Color vision depends initially on the presence, in the retina, of receptors containing photopigments of different spectral sensitivities. The information about wavelength of light is usually then transmitted by spectrally opponent cells that are excited by part of the spectrum and inhibited by the rest. A major problem is to identify and describe the spectra of the receptors contributing to the responses of these later stages. A novel analytical method has been developed which permits us to decompose any spectral sensitivity function into its component photopigment spectra. One of the aims is also to specify the transformations and interactions between the transduction of light by the receptors and the responses at some later stage. Responses will be recorded from ganglion cells and horizontal cells in the goldfish retina. This species has excellent color vision and can serve as a model for vertebrate vision. Thus, conclusions should be widely applicable. However, the work is important not only for the specific understanding of the goldfish retina, or even a general vertebrate retina; for example, the proposed techniques for spectral analysis are in principle general and have potentially direct applications to humans--if the methods prove valid, it should be possible to take any spectral sensitivity function (not just a physiological one) and to identify it from the spectra of the underlying photopigments.